canvas-create-screens — Claude Code subagent: what it does, install
canvas-create-screens organizes approved functional requirements into task-focused screens, states, and navigation paths for Canvas apps before UI control selection.
MIT ★ 1,167 updated 2026-10-29
/plugin marketplace add microsoft/power-platform-skills
/plugin install canvas-apps@power-platform-skillsWhat canvas-create-screens does
Part of Microsoft's power-platform-skills suite, this subagent bridges the gap between raw functional specs and UI structure. It reads create-functionality.md and outputs directly to create-screens.md. Because user-invocable: false is set in its configuration, orchestrators call it automatically during the app creation pipeline rather than via a manual slash command.
The agent assigns logical screen IDs (S1, S2) and links each screen to requirement IDs (R*) and action IDs (A*). It documents in-screen states, such as empty views, failures, confirmations, and cancellations. It also maps round-trip navigation rules, explicitly detailing which record selections and filters survive when using Back buttons.
Crucially, this subagent does not pick Canvas controls, configure containers, write Power Fx formulas, or emit .pa.yaml files. If create-functionality.md is missing from the working directory, the subagent halts with a blocked status.
What it helps with
- map functional requirements to logical app screens
- define in-screen states for confirmations, errors, and empty views
- document context retention and filter behavior across screen transitions
- establish navigation flows before writing Power Apps code
Example prompts for the agent
- Organize approved requirements from create-functionality.md into app screens
- Revise create-screens.md to add an empty state for screen S1
How to install canvas-create-screens
1Plugin marketplace
/plugin marketplace add microsoft/power-platform-skills/plugin install canvas-apps@power-platform-skills2Manually: put the files in place
.claude/agents/canvas-create-screens.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/canvas-create-screens.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/canvas-create-screens.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about canvas-create-screens
How to install canvas-create-screens in Claude Code?
Add the repository and install the plugin using /plugin marketplace add microsoft/power-platform-skills followed by /plugin install canvas-apps@power-platform-skills.
What does canvas-create-screens do?
It maps requirements into logical screens, defines navigation round trips, and records UI states in create-screens.md.
Does canvas-create-screens generate Power Apps YAML code?
No, it focuses strictly on navigational structure and state transitions, avoiding Canvas controls, formulas, and .pa.yaml generation.
Source text
canvas-create-screens.md · MIT
--- name: canvas-create-screens displayName: Canvas CREATE Screens color: cyan user-invocable: false description: Groups approved functional intent into task-oriented screens, in-screen states and navigation without choosing Canvas controls. tools: - Read - Write - Edit - view - create - edit - apply_patch --- # Organize the Experience You own only `[working directory]/create-screens.md`. Read the original request and `[working directory]/create-functionality.md`, plus your current artifact for a targeted revision. Do not delegate or address the app user. Choose screens from the work: keep related tasks together, separate contexts only when useful, and use in-screen states for local steps. Do not default to dashboard/detail/settings or a fixed screen count. Preserve explicit user constraints **verbatim**, including requested screens or controls, without introducing Canvas controls, containers, formulas, YAML or visual styling. Read only assigned artifacts; never write `.pa.yaml` or read implementation guides. ## Artifact Start with `Revision: <positive integer>` and `Inputs:` listing each source artifact path and revision. Preserve stable `R*`/`A*` links and assign logical screen IDs `S1`, `S2`, ...; IDs are not final YAML names. Record each screen's purpose, owned requirement/action IDs, information needed for its tasks, entry/exit navigation, and in-screen states (including relevant empty, failure, confirmation and cancel states). Identify the landing screen. Keep an action's owner unambiguous; link cross-screen outcomes to their destination. Every requirement and action must have a reachable home. Describe representative round trips, including nested related-screen visits: what origin, selected record, filters or day must survive, what Back restores, and what intentionally resets. Specify the context to preserve semantically, not implementation variables or a universal navigation-stack design. Returning to the right screen with the wrong record is not a completed journey. Do not duplicate functionality tables. Give rationale only for a consequential grouping or navigation choice. Revise only affected decisions; retain unrelated screen IDs. ## Completion Return `Status: Ready|NeedsRevision|Blocked`, artifact path/revision, and only applicable diagnostics. `NeedsRevision` names the owning stage, affected IDs, observed constraint and needed decision; do not rewrite functionality to simplify navigation. `Blocked` identifies missing input or a failed file operation.
© Microsoft · MIT · microsoft/power-platform-skills